Two-stage generative adversarial networks for document image binarization with color noise and background removal

20 Oct 2020  ·  Sungho Suh, Jihun Kim, Paul Lukowicz, Yong Oh Lee ·

Document image enhancement and binarization methods are often used to improve the accuracy and efficiency of document image analysis tasks such as text recognition. Traditional non-machine-learning methods are constructed on low-level features in an unsupervised manner but have difficulty with binarization on documents with severely degraded backgrounds. Convolutional neural network-based methods focus only on grayscale images and on local textual features. In this paper, we propose a two-stage color document image enhancement and binarization method using generative adversarial neural networks. In the first stage, four color-independent adversarial networks are trained to extract color foreground information from an input image for document image enhancement. In the second stage, two independent adversarial networks with global and local features are trained for image binarization of documents of variable size. For the adversarial neural networks, we formulate loss functions between a discriminator and generators having an encoder-decoder structure. Experimental results show that the proposed method achieves better performance than many classical and state-of-the-art algorithms over the Document Image Binarization Contest (DIBCO) datasets, the LRDE Document Binarization Dataset (LRDE DBD), and our shipping label image dataset. We plan to release the shipping label dataset as well as our implementation code at github.com/opensuh/DocumentBinarization/.

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Binarization DIBCO 2011 Two-stage generative adversarial networks for binarization of color document images PSNR 20.22 # 7
F-Measure 93.57 # 7
DRD 1.99 # 7
Pseudo-F-measure 95.93 # 5
Binarization DIBCO 2013 Two-stage generative adversarial networks for binarization of color document images F-Measure 95.01 # 5
Pseudo-F-measure 96.49 # 5
PSNR 21.99 # 5
DRD 1.76 # 5
Binarization H-DIBCO 2014 Two-stage generative adversarial networks for binarization of color document images F-Measure 96.36 # 4
Pseudo-F-measure 97.87 # 4
PSNR 21.96 # 4
DRD 1.07 # 4
Binarization H-DIBCO 2016 Two-stage generative adversarial networks for binarization of color document images F-Measure 92.24 # 4
PSNR 19.93 # 4
DRD 2.77 # 3
Pseudo-F-measure 95.95 # 3
Binarization LRDE DBD Two-stage generative adversarial networks for binarization of color document images F-Measure 98.01 # 1
Pseudo-F-measure 98.33 # 1
PSNR 27.79 # 1
DRD 0.73 # 1

Methods


No methods listed for this paper. Add relevant methods here